Selection wrappers: improve/optimize MergedSelection & TupleSelection

Improve/optimize selection wrappers:

1. MergedSelection is rewritten to generate much more optimal structure of algorithms/sequences. It makes it much more efficient, since
   the merging/filtering algorithm is invoked only when it is really needed.
2. TupleSelection is a class now, that allows more decorations to be used, in particular it exposes the most important methods/properties of underlying algorithm, that makes configuration easier

my_selection = ...
    
from PhysConf.Selections import TupleSelection 
my_tuple = TupleSelection(
    'TUPLE'          , ## unique name 
     my_selection    , ## input selection
    Decay = "[Lambda_b0 -> ^(Lambda_c+ -> ^p+ ^K- ^pi+) ^pi-]CC" , 
    Branches = {
    "Lb":"[Lambda_b0 -> (Lambda_c+ -> p+ K- pi+) pi-]CC",
    "Lc":"[Lambda_b0 -> ^(Lambda_c+ -> p+ K- pi+) pi-]CC",
    } ,
		MoreTools = [ 
    'TupleToolKinematic'  ,
    'TupleToolPid'        ,
    'TupleToolANNPID'     ,
    'TupleToolGeometry'   ,
    "TupleToolPropertime" ,
		] 
  	  )

## update tool list 
my_tuple.ToolList += [
    "TupleToolTrackInfo"  ,
    "TupleToolTrigger"    , 
    "TupleToolPrimaries"
    ]
from Configurables import  ( TupleToolTISTOS          ,
                             TupleToolDecay           ,
                             TupleToolPrimaries       ,
                             TupleToolDecayTreeFitter )

## update tool list
my_tuple.ToolList+=[ "TupleToolTISTOS" ]

## access the tool by name
tool = Lb_tuple.TupleToolTISTOS
tool.Verbose=True
tool.VerboseHlt1=True
tool.TriggerList = [
    "L0MuonDecision", "L0DiMuonDecision", "L0HadronDecision",
    "Hlt1TrackMVADecision" , "Hlt1TrackMVALooseDecision",
    "Hlt1TwoTrackMVADecision", "Hlt1TwoTrackMVALooseDecision",
    ]

## access the branch:
Lb = Lb_tuple.Lb
Lb.ToolList += ["TupleToolDecayTreeFitter/PVFit"]
Lb.addTool(TupleToolDecayTreeFitter("PVFit"))
Lb.PVFit.Verbose = True
Lb.PVFit.UpdateDaughters = True